Build to Suit Data Center: Structure, Economics and Risk
A build to suit data center is a facility constructed to a named tenant’s specification, under a lease agreed before the building is finished. The developer funds and delivers. The tenant commits to a long term at a rent that returns the developer’s cost plus a margin.
It is the structure most developers would prefer, because it removes the question that makes speculative development risky. The tenant is known before the capital is committed, which is why a data center developer with a credible requirement in hand will almost always choose this route.
The developer funds and delivers, and the tenant commits to 10 to 20 years at a rent that typically returns delivered cost plus a margin. It removes leasing risk and replaces it with construction and delivery risk, produces the highest rent per kilowatt of the three structures available to a developer, and requires the most capital. Two things decide whether one is achievable: the tenant’s covenant, because the lease is what the developer’s financing is secured against, and the density specification, because a requirement at 30 kilowatts a rack and one at 150 produce genuinely different buildings.
Metro Colo Advisory advises developers and investors on whether a site can support a build to suit, which tenants would consider it, and what the arrangement returns. This page covers how the structure works and what determines whether anyone signs. If you hold a site and are weighing this against building speculatively, our data center consulting practice exists for that question.
Engagements are project-based and fixed-fee, typically from $40,000 to $195,000 depending on scope.
How a Build to Suit Data Center Differs From the Alternatives
Three structures are available to a developer holding a site with power. They differ in when the tenant appears and who carries the risk in between.
| Build to Suit | Powered Shell | Turnkey Data Center | |
|---|---|---|---|
| When the tenant commits | Before construction | Before or during | After completion |
| Who specifies the building | The tenant | The developer, to market standards | The developer, to a guess |
| Who funds the fit out | Developer, recovered in rent | Tenant | Developer |
| Developer capital at risk | Committed against a signed lease | Lower, but speculative | Highest and speculative |
| Lease term | 10 to 20 years | 10 to 20 years | Shorter, and variable |
| Rent per kilowatt | Highest of the three | Lowest | Market at the time of lease-up |
| Main risk | Construction cost and delivery date | Finding a tenant at all | Lease-up duration |
The trade is straightforward. A build to suit data center removes leasing risk and replaces it with delivery risk. A powered shell keeps capital exposure low but leaves the developer holding a building nobody has committed to. A speculative turnkey data center carries both, which is why very few get built without a tenant in view.
Working out which of the three a specific site should pursue is where our engagements usually start, because the specification, the cooling architecture and the electrical design all follow from that decision.
Not sure which structure fits your site?
The Commercial Structure
A data center build to suit typically runs through five stages.
Letter of intent
The tenant states the requirement in megawatts, density, timeline and location. The developer confirms the site can serve it. Neither party is committed, but the specification starts to firm up.
Specification and cost
The tenant’s technical requirements are translated into a design and a construction budget. This is where most build to suit arrangements succeed or fail, because the tenant wants a facility built for their hardware and the developer wants a building that has value if the tenant leaves in fifteen years.
Lease negotiation
Rent is usually set as a return on total project cost rather than as a market rate, since there is no comparable market for a purpose-built facility. Typical structures reference a yield on the developer’s delivered cost, with mechanisms for cost overruns, delivery delays and specification changes.
Construction
The developer builds. The tenant inspects. Change orders are the most common source of dispute, because a tenant whose hardware roadmap moves will want the building to move with it.
Delivery and commencement
Rent starts on delivery, or on a date agreed regardless of delivery, depending on how the lease allocates delay risk.
How Long Does It Take to Build a Data Center
The honest answer is that construction is rarely the long pole. A build to suit typically runs 18 to 30 months from signed lease to delivery, but the work that determines whether that date holds happens before the lease is signed.
Data center interconnection is the first constraint. A standard interconnection agreement runs roughly three years from initial request to final agreement, and delays past six years are not unusual. Across major US markets the average wait from application to commercial operation is around five years. A position in the interconnection queue is not power, and the substation work behind it carries its own timeline.
Equipment lead times are the second. Substation transformers currently run past 160 weeks. Generator step-up transformers run 144 weeks. Switchgear runs 52 to 78 weeks and UPS systems 30 to 40. Long-lead equipment frequently has to be ordered before the lease is signed, which puts the developer at risk on equipment they cannot use if the tenant walks.
Entitlement is the third and the least predictable. Community opposition has blocked an estimated $18 billion in US data center projects and delayed a further $46 billion by more than two years. In Home Rule jurisdictions a planning commission recommendation can be reversed by a council vote.
Which is why the commercial operation date committed in a build to suit lease is the hardest number in the agreement. It has to be achievable on the day the lease is signed, against an interconnection queue and an equipment supply chain the developer does not control.
A realistic sequence
- Site control, entitlement and interconnection application: 12 to 36 months, often concurrent
- Design and specification lock: 3 to 6 months
- Long-lead equipment procurement: ordered during design, arriving 12 to 40 months later
- Construction: 12 to 24 months
- Data center commissioning, levels 1 through 5: 2 to 4 months
What commissioning actually means
Data center commissioning is the process of proving the facility works under load before anything is installed in it. It runs in five levels. Factory testing of individual components. Delivery inspection on site. Pre-functional checks that each system is installed correctly. Functional testing of each system on its own. And integrated systems testing, where the whole facility is run under simulated load with utility power pulled to confirm it behaves as designed.
A lease that commits to a delivery date is committing to the end of level 5, not to the end of construction. That distinction is worth two to four months and it is frequently missed in negotiation.
A site with entitlement and interconnection already secured can deliver in two to three years. A site starting from raw land is realistically five or more. That gap is most of what determines whether a build to suit is achievable at all, and it is why a developer with a credible tenant and a shovel-ready site holds something scarce.
What a Tenant Specifies
Build to suit requirements are more detailed than a colocation contract, because the tenant is designing the building rather than renting space in one.
Power. Contracted capacity with an energization date, redundancy configuration specified against a tier standard, and the path to expansion. Utility indications are not sufficient at this stage.
Density. Kilowatts per rack, which determines almost everything else. The market average reached about 27 kilowatts in 2026, up from 16 the year before, while rack-scale platforms draw 140 to 160 and the published roadmap runs past 600.
Cooling architecture. Whether facility water reaches the white space. Air cooling has a hard ceiling near 30 to 35 kilowatts per rack, so any requirement above that specifies liquid cooling at the design stage rather than as a retrofit.
Floor loading. Design live load in pounds per square foot. ANSI/TIA-942 sets 150 as a minimum and 250 as recommended. Rack-scale deployments weigh roughly 1.5 tons per rack, and Intel recommends 350 psf for high density data centers.
Connectivity. Carrier count, and more importantly physical route diversity. Two providers on one path is a single point of failure however it reads on a marketing sheet. The same logic applies earlier, when a tenant is choosing between markets rather than buildings.
Expansion rights. Whether the tenant can take additional capacity, on what terms, and by when. Expansion rights that cannot be met are worse than no expansion rights, because they create an obligation the developer cannot discharge.
Establishing whether a site can meet a specification like that is the first phase of a data center consulting engagement. The output is a density ceiling by zone rather than a single figure, and it determines which tenants are worth approaching at all.
How Density Changes the Building
The following is illustrative, using round numbers rather than any specific project, but it shows why a tenant’s density figure matters more than any other number in the specification.
Take a requirement for 7 megawatts of IT load. At 30 kilowatts per rack that is roughly 230 racks. At 150 kilowatts per rack it is roughly 47.
Same power. Same rent, since the lease prices per kilowatt. But the second building needs about a fifth of the white space, four to five times the structural floor loading, a cooling architecture that delivers liquid to every rack rather than air to a room, and electrical distribution designed around a completely different circuit configuration.
They are not the same building with different equipment in it. They are different buildings.
Which is why a tenant who firms up density late in the specification process has effectively asked for a redesign, and why a developer who agrees a build to suit before the density is fixed has agreed to an unknown cost. The most common version of this is a tenant specifying for the hardware they run today while their roadmap moves to something denser within the lease term.
Getting that number settled early, and building in the headroom to accommodate the next generation, is worth more than any other clause in the agreement. Building toward the wrong density is not a rounding error. On a project of this scale it is a mistake measured in millions, and it is decided months before anyone notices.
The Economics
Build to suit produces the highest rent per kilowatt of the three structures, because the developer is funding the entire facility and recovering it through the lease.
Published benchmarks put turnkey data center construction at roughly 8 to 12 million dollars per megawatt, and 15 to 20 million or more per megawatt for facilities designed for rack-scale AI density. Electrical infrastructure alone accounts for 40 to 45 percent of that. A build to suit developer carries all of it.
Against that, the return is contracted. Rent is typically set as a yield on delivered cost, which means the developer knows the return before breaking ground rather than discovering it during lease-up. Digital Realty guided to an 11.5 percent average expected stabilized yield on its development pipeline in mid-2026, which is a useful reference for what institutional capital targets on this kind of work.
The other advantage is the exit. A completed facility with a fifteen year lease to a creditworthy tenant trades at a materially different capitalization rate than a speculative building, and that spread is where most of the development profit sits.
Yield on cost is the figure most developers want and the hardest to source, because the rent in a build to suit is negotiated rather than observed. We model it in ranges with the assumptions stated and sensitivity shown on construction cost, delivery timing and tenant covenant.
When a Build to Suit Data Center Is the Right Structure
It fits when a credible tenant exists and is prepared to commit before construction. That is the whole basis of the structure and there is no version of it without one.
It fits when the site can meet a specification that a specific tenant actually wants, rather than a generic one. Build to suit is not speculative development with a tenant attached; the building is designed around the requirement.
It fits when the data center developer can carry construction risk, including cost escalation on electrical and mechanical equipment where lead times have extended significantly.
It fits less well when the tenant’s covenant will not support a fifteen year lease, when the requirement is small enough that the specification cost cannot be amortized, or when the tenant’s hardware roadmap is moving faster than the building can be designed. That last one has become common, since colocation terms run five to ten years while the hardware envelope moves every twelve months.
Credit Is the Constraint Nobody Plans For
A build to suit lease is a fifteen or twenty year receivable, and it is what the developer’s financing is secured against. Which makes the tenant’s covenant as important as the specification.
Developers and their lenders underwrite the entity signing the lease, not the business behind it. A requirement from a well-funded company operating through an unrated subsidiary is a different proposition to a lender than the parent name suggests, and the gap between those two things has ended more build to suit conversations than any technical issue.
Where a covenant sits below investment grade, the structures that bridge it are a larger deposit, a letter of credit, a prepaid term, or a parent guarantee. Establishing which of those a tenant can offer is worth doing early rather than after a specification has been agreed.
What Gets Locked at Design
Three decisions cannot be reversed economically once construction proceeds.
Floor loading, which sets the density ceiling. Slab to slab clear height, which governs whether overhead distribution and containment fit. And whether the cooling loop reaches the white space or stops at the plant room.
In a build to suit these are agreed with the tenant rather than guessed, which is the structure’s main advantage. The risk is different: a tenant who specifies for today’s hardware may find the building constrains them by year five, and a developer who agrees to an over-specified design carries cost that will not be recovered if the tenant leaves.
The cost of changing a specification on a drawing is a fraction of changing it once the slab is poured
Still pre-construction? That is the moment this analysis is worth most.
Common Mistakes
Treating a letter of intent as a commitment.
Specification work is expensive, and a tenant who walks after the design phase leaves the developer with drawings for a building nobody wants.
Designing so tightly that the building has no residual value.
A facility built entirely around one tenant’s hardware is difficult to re-let. The best build to suit designs meet the requirement while remaining serviceable to a broader tenant set.
Underwriting the brand rather than the entity.
The name on the requirement and the name on the lease are frequently different, and only one of them matters to a lender.
Agreeing expansion rights without contracted power to support them.
Utility studies for additional capacity can run one to two years, and an expansion right that cannot be met is a liability rather than a feature.
Assuming carrier presence equals route diversity.
Two providers on the same physical path is a single point of failure, and a tenant with uptime commitments will ask for route maps rather than carrier lists.
Each of these is cheap to correct before a lease is signed and expensive afterwards. They are also the kind of thing a real estate advisor and an engineering firm can both miss, because neither is asked to underwrite the tenant.
Frequently Asked Questions
What is a build to suit data center?
A facility constructed to a named tenant’s specification under a lease agreed before the building is finished. The developer funds and delivers the facility; the tenant commits to a long term, usually 10 to 20 years, at a rent that returns the developer’s cost plus a margin.
What is the difference between build to suit and powered shell?
Build to suit has a tenant before construction and the building is designed around their requirement, with the developer funding the full fit out. A powered shell delivers the building, power and cooling plant with the tenant funding the interior, and can be leased speculatively. Build to suit produces higher rent per kilowatt and requires substantially more developer capital.
How is rent set in a data center build to suit?
Usually as a yield on the developer’s total delivered cost rather than as a market rate, because there is no comparable market for a purpose-built facility. The lease will normally include mechanisms for cost overruns, delivery delays and specification changes.
How long does a build to suit data center take to deliver?
The construction period depends on the site, but the binding constraint is usually power rather than building. Where additional utility capacity is required, transmission studies can run one to two years before construction timelines are even relevant.
Who takes the risk in a build to suit?
The developer carries construction cost and delivery date risk. The tenant carries the risk of committing to a facility for fifteen or twenty years while their own requirements change. The lease allocates delay and cost overrun risk between them, and that allocation is usually the most negotiated part of the agreement.
What credit does a tenant need for a build to suit?
Stronger than for a colocation contract, because the lease is what the developer’s financing is secured against. Where a covenant sits below investment grade, the bridging structures are a larger deposit, a letter of credit, a prepaid term, or a parent guarantee.
Can a build to suit be converted to a shell if the tenant walks?
Sometimes, and it depends entirely on how tightly the building was specified. A facility designed around one tenant’s hardware is difficult to re-let, which is why residual value should be considered during the specification stage rather than after.
What This Page Cannot Tell You
Everything above is general. Four things are specific to a site, and none of them can be answered from a page.
- Whether your site can meet a specific tenant’s specification. That comes from the structural general notes, the electrical single line diagram and the mechanical drawings, measured against what a named tenant actually requires.
- Which tenants would consider a build to suit at your site. That requires knowing current requirements in your market, their density, their timeline, and whether the covenant behind them will support a long lease. None of it is published.
- What rent would return your cost. Build to suit rent is negotiated rather than observed, and it depends on your delivered cost, your required return and the tenant’s alternatives at the time.
- What any of it returns. Yield on cost depends on your basis in the land, what you have already committed, your capital position and your hurdle rate. Two identical buildings produce very different numbers depending on how the owner got there.
Those four are what a consulting engagement produces, and they are the reason a page like this can only take the question so far.
Working Out Whether a Build to Suit Is Achievable
Metro Colo Advisory advises developers and investors on exactly this question. A full engagement covers six things.
- Building envelope. What the site can serve and at what density, given power, floor loading, cooling architecture, clear height and area for mechanical plant.
- Operator and tenant fit. Which operators and end users match the site and the market, with the reasoning stated for each.
- Market context. Competitive supply in the region, and what tenants at each density tier actually require.
- Yield on cost. Capital, revenue, operating burden and exit value across build to suit, powered shell and speculative development, modeled against a defined budget.
- Connectivity. Carrier position, physical route diversity, and the options for improving it.
- Specification and recommendation. Which path the analysis supports, what the building needs to be to capture it, and the sequence against your construction timeline.
Engagements typically run from $40,000 to $195,000 depending on scope, with monthly retainers available for ongoing work.
We are independent. We design nothing and operate nothing, and our consulting engagements are paid by the client rather than by any provider. Separately, we place enterprise and infrastructure requirements into colocation facilities nationally, which is how we know what tenants at each density tier actually ask for rather than what they say in a survey.
